A user asks readers to consider what either answer would imply for how things will go.
The post poses a question about whether intelligence per active parameter has a limit, then invites readers to think through the implications of either answer.
30.4K
2 Sources, first seen ago
—
Not ranked yet
—
Not ranked yet
@willcbask yourself whether you feel that intelligence per active parameter is bounded or unbounded and then consider what that would imply about how things will go
@sytelus@willcb Bound should be per flop, not parameter. For given parameters, you can get ton of flops by simply increasing context. The upper bound per flop could be very high and we are no where close to hit it.@willcbask yourself whether you feel that intelligence per active parameter is bounded or unbounded and then consider what that would imply about how things will go@sytelus@willcb Bound should be per flop, not parameter. For given parameters, you can get ton of flops by simply increasing context. The upper bound per flop could be very high and we are no where close to hit it.@willcbask yourself whether you feel that intelligence per active parameter is bounded or unbounded and then consider what that would imply about how things will go@sytelus@willcb Bound should be per flop, not parameter. For given parameters, you can get ton of flops by simply increasing context. The upper bound per flop could be very high and we are no where close to hit it.@willcbask yourself whether you feel that intelligence per active parameter is bounded or unbounded and then consider what that would imply about how things will go@sytelus@willcb Bound should be per flop, not parameter. For given parameters, you can get ton of flops by simply increasing context. The upper bound per flop could be very high and we are no where close to hit it.
